An open-weight AI model is one whose trained weights are published for anyone to download and run, which differs from open source as the Open Source Initiative defines it, with freedoms to use, study, modify and share the system. The initiative describes open weights as "the final weights and biases of a trained neural network" and says they differ from open source AI because they do not include the training code or the training dataset. The gpt-oss-120b card says that model can "run on a single 80GB GPU," and the license attached to the files sets the terms for using them.
How the terms differ
| Term | Who defines it | What it requires |
|---|---|---|
| Open weights | Described by the Open Source Initiative; also used in model cards | The final trained weights are shared; the training code and training dataset are not included |
| Widely available model weights | NTIA, July 30, 2024 | Weights "open to the public by allowing users to download these weights from the Internet or through other mechanisms" |
| Open Source AI | Open Source Initiative, definition 1.0 | The freedoms to use, study, modify and share the system |
The Open Source Initiative's page words the four freedoms this way: "Use the system for any purpose and without having to ask for permission," "Study how the system works and understand how its results were created," "Modify the system for any purpose, including to change its output," and "Share the system for others to use with or without modifications, for any purpose." It says each freedom depends on access to "the preferred form to make modifications to the system, and to the means to use it."